Comparison of alternative loss modules in the IHACRES Model: an application to 7 catchments in Wales

Abstract

The IHACRES model has been successfully applied to many catchments across the UK including regionalization studies (e.g. Sefton and Howarth, 1998; Littlewood, 2003). However, while the previous formulation of the non-linear module performs well in calibration and simulation, the module has some identified deficiencies in the representation of the impacts of climate change on volume of streamflow. The recently developed Catchment Moisture Deficit (CMD) version of the non-linear module (Croke and Jakeman, 2004) has a stronger physical basis than the previous formulation of the IHACRES non-linear loss module, which potentially enhances the performance in regionalization studies. The ultimate aim of this study is to apply the model to 60 catchments in England and Wales and examine potential regionalization strategies using the CMD model. The previous IHACRES model has already been applied to a regionalization study based on these catchments, and a comparison of the two results will be used to evaluate the strengths and weaknesses of the two versions of IHACRES for such studies. This paper describes initial tests on the performance of the revised IHACRES model on 7 catchments in Wales. The catchments span a range of areas from 129 to 1480 km2 with most having a high runoff coefficient of greater than 0.7, with the remaining catchment located in east Wales, and has a runoff coefficient of 0.44 (Littlewood, 2003). Using cross correlation analysis, one of the 7 catchments (Teme) was found to have no seasonality in the rainfall distribution and only a slight seasonality in the streamflow response. In comparison the remaining 6 catchments all had strong seasonal variation in the streamflow driven by a slight seasonality in the rainfall. In addition, all 6 catchments had very similar streamflowrainfall cross correlation functions, indicating that a similar model structure should apply to all 6 catchments, and that there is a strong potential for regionalization. Based on deconvolution of the cross correlation functions, the delay between rainfall and the peak of the catchment response function is between 0 and 1 days. The centre of gravity of the response function is a measure of both the delay in the peak as well as the rate of decay from the peak, and has a significant correlation with catchment area. The CMD version of the IHACRES gave better performance than the non-linear module of Jakeman and Hornberger (1993) for the Teme catchment (RNS2 = 0.70 compared with 0.65 found by Littlewood, 2003). However, for five of the remaining six catchments, the CMD module performed slightly worse than the previous version of the module. While the number of catchments included in this study is too small to provide a conclusive analysis of the performance of the CMD module, the results suggest that the CMD module performs better than the previous version on larger catchments.